Douyin Transcripts Scraper | 抖音 Subtitles & Translations in 50+ Languages (2026)

Timestamped transcripts & .srt subtitles for any 抖音 video, any length — Mandarin → 50 languages · Creator + post metadata · No duration caps · Frontier AI quality

Zen Studio · 抖音 + 西瓜视频 (Xigua)   •  Full-stack China short-video data: profiles, posts, transcripts, subtitles
Douyin Transcripts ➤ You are here	Douyin Profiles 博主 followers, posts, hashtags	Xigua Search 西瓜视频 keyword + comments	Xigua Detail 70+ fields + MP4 下载

Copy to your AI assistant

zen-studio/douyin-transcripts-scraper on Apify. Returns timestamped transcripts and optional .srt subtitles for any 抖音 (Douyin) video, in Mandarin or any of 50 translated languages. Call ApifyClient("TOKEN").actor("zen-studio/douyin-transcripts-scraper").call(run_input={...}), then client.dataset(run["defaultDatasetId"]).list_items().items. Full spec: GET https://api.apify.com/v2/acts/zen-studio~douyin-transcripts-scraper/builds/default (Bearer TOKEN) → inputSchema, actorDefinition.storages.dataset, readme. Token: https://console.apify.com/account/integrations

Key Features

• Frontier-model transcription quality — best-in-class Mandarin speech-to-text paired with state-of-the-art LLM translation. Two-tier translation pipeline self-escalates to a stronger model when needed for line-aligned accuracy.
• 51 languages — 中文 (zh, no translation) plus 50 translation targets covering ~95% of global speakers. Top picks first (English, Spanish, Hindi, Arabic, Portuguese, Russian, Japanese, German, French, Korean, …) then the long tail (Bulgarian, Catalan, Czech, Hebrew, Swedish, Tamil, …). Full table below.
• Timestamped per-cue segments — every line of the transcript carries start / end floats so you can sync to player timelines, build interactive transcripts, or edit clips around specific phrases.
• Optional .srt subtitle export — flip one switch and get a SubRip (.srt) subtitle file ready to upload to YouTube, Instagram, TikTok, Premiere, CapCut, Final Cut, DaVinci Resolve. Pay only when you ask for it.
• Bonus: keep the MP4 + cover, free — flip shouldDownloadVideos and the original MP4 lands in your key-value store alongside the transcript. The actor already streams the video to extract audio, so saving a copy costs nothing extra. Same toggle for cover images. No need to also run the Douyin Video Scraper.
• Full creator + post metadata — 60+ fields per video: author profile, statistics, hashtags, music, chapters, video tags, share URL, cover image, and more. Same shape as the Douyin Video Scraper, with the transcript layered on top.
• Free tier — 5 minutes of audio lifetime, no credit card. Long videos get a partial transcript so you can preview the output before committing.
• Robust URL parsing — paste a desktop URL, a v.douyin.com/... short link, a ?modal_id=... web-feed URL, the entire mobile share blob (复制此链接 ...), or just a numeric aweme ID. All work.

How to Get 抖音 Video Transcripts

Basic — original Chinese transcript

{
  "videoUrls": [
    "https://www.douyin.com/video/7534679152504376595"
  ],
  "targetLanguage": "zh"
}

English subtitles for a Douyin video

{
  "videoUrls": [
    "https://www.douyin.com/video/7534679152504376595"
  ],
  "targetLanguage": "en",
  "outputSrt": true
}

Bulk translation — German subtitles for a batch

{
  "videoUrls": [
    "https://www.douyin.com/video/7534679152504376595",
    "https://www.douyin.com/video/7627491832754228514",
    "https://v.douyin.com/abc123/",
    "8.79 0bH:/ # 看看这个 https://v.douyin.com/xyz999/ 复制此链接"
  ],
  "targetLanguage": "de",
  "outputSrt": true
}

Archive .srt files across many runs

{
  "videoUrls": ["7534679152504376595"],
  "targetLanguage": "en",
  "outputSrt": true,
  "srtKvStoreName": "douyin-subtitles-archive"
}

The default key-value store is wiped 7 days after a run. Use srtKvStoreName to drop your subtitle files into a named store that lives forever and accumulates across runs.

Input Parameters

Parameter	Type	Default	Description
videoUrls	array of strings	required	One or more 抖音 video URLs. Most URL formats accepted (see below). Max 5 per run.
targetLanguage	string (enum)	"zh"	Output language. zh returns the original Chinese transcript with no translation cost. Any other code triggers an LLM translation pass.
outputSrt	boolean	false	When true, also generates a SubRip (.srt) subtitle file and includes srtUrl on the row.
srtKvStoreName	string	""	Optional named key-value store for .srt files. Lowercase letters, digits, and dashes only. Empty = the run's default store.

Supported URL formats

All of these resolve to the same video:

• https://www.douyin.com/video/7534679152504376595 — desktop URL
• https://www.iesdouyin.com/share/video/7534679152504376595/ — share URL
• https://v.douyin.com/<short>/ — mobile app share link
• https://www.douyin.com/jingxuan/sports?modal_id=7534679152504376595 — web-feed modal
• https://www.douyin.com/note/7534679152504376595 — image-post URL
• 7534679152504376595 — bare numeric aweme ID
• 8.79 0bH:/ # 看看这个 https://v.douyin.com/abc/ 复制此链接 ... — full mobile share blob (URL extracted)

Supported languages (targetLanguage)

zh returns the original Mandarin transcript with no translation pass (lowest cost). Any other code triggers an LLM translation. The list is curated to languages where the model produces natural, conversational output — top demand first, then alphabetical.

Code	Language	Code	Language
zh	中文 — Chinese (no translation)	en	English
es	Español — Spanish	hi	हिन्दी — Hindi
ar	العربية — Arabic	pt	Português — Portuguese
ru	Русский — Russian	ja	日本語 — Japanese
de	Deutsch — German	fr	Français — French
ko	한국어 — Korean	id	Bahasa Indonesia — Indonesian
tr	Türkçe — Turkish	it	Italiano — Italian
vi	Tiếng Việt — Vietnamese	th	ภาษาไทย — Thai
pl	Polski — Polish	nl	Nederlands — Dutch
bn	বাংলা — Bengali	ur	اردو — Urdu
fa	فارسی — Persian	uk	Українська — Ukrainian
ms	Bahasa Melayu — Malay	tl	Filipino — Filipino (Tagalog)
bg	Български — Bulgarian	ca	Català — Catalan
cs	Čeština — Czech	da	Dansk — Danish
el	Ελληνικά — Greek	et	Eesti — Estonian
fi	Suomi — Finnish	gu	ગુજરાતી — Gujarati
he	עברית — Hebrew	hr	Hrvatski — Croatian
hu	Magyar — Hungarian	is	Íslenska — Icelandic
kn	ಕನ್ನಡ — Kannada	lt	Lietuvių — Lithuanian
lv	Latviešu — Latvian	ml	മലയാളം — Malayalam
mr	मराठी — Marathi	nb	Norsk bokmål — Norwegian (Bokmål)
pa	ਪੰਜਾਬੀ — Punjabi	ro	Română — Romanian
sk	Slovenčina — Slovak	sl	Slovenščina — Slovenian
sr	Српски — Serbian	sv	Svenska — Swedish
sw	Kiswahili — Swahili	ta	தமிழ் — Tamil
te	తెలుగు — Telugu

What Data Can You Extract from 抖音 Videos?

Every dataset row pairs the transcript and subtitle data with the same rich metadata shape used by the Douyin Video Scraper. 60+ top-level fields per video.

Transcript-specific fields

• transcript — full plain-text transcript in the chosen language
• segments — array of {start, end, text} cues with float second timing
• language — output language code (zh, en, de, ...)
• srtUrl — public link to the .srt file in your key-value store (only when outputSrt: true)
• duration — total video duration in seconds (float)
• truncatedFromSeconds — set on free-tier rows where the audio was truncated to fit the quota
• mp4Url — direct link to the saved MP4 in your key-value store. Populated when shouldDownloadVideos: true, null otherwise. Same URL as videoFile.kvUrl, lifted to the top level so it's clickable in the Apify Console and easy to pipe into a download script.
• coverImageUrl — direct link to the saved cover JPEG. Populated when shouldDownloadCovers: true, null otherwise. Lifted from coverFile.kvUrl.
• videoFile — full metadata block for the saved MP4 (kvUrl, kvStoreKey, kvStoreId, sizeBytes, mimeType, sourceUrl) when shouldDownloadVideos: true, else null. Use this when you need the file size or the original source CDN URL alongside the saved location.
• coverFile — same shape as videoFile for the cover JPEG. null when the toggle is off.

Metadata fields

• Identity & URLs: id, groupId, url, shareUrl, videoUrl, audioUrl, inputUrl
• Type & flags: type, awemeType, mediaType, horizontalType, plus 14 boolean flags (isPinned, isAd, isPgc, isStory, isVr, ...)
• Caption variants: text (full description), caption, itemTitle, previewTitle, descLanguage
• Time & locality: createTime, createDate, region, city, cityCode
• Nested rich blocks:
  • authorMeta — name, sec_uid, verified status, total likes received, avatars (5 sizes), signature, region, demographics (gender/language/userAge/birthday/constellation), commerce flags, school, cross-platform IDs (42 fields)
  • videoMeta — cover image, dimensions, ratio, bit-rate variants, watermark flag, CDN expiry
  • musicMeta — track title, author, original-sound flag, play URL, cover art
  • statistics — diggCount, commentCount, shareCount, collectCount, downloadCount, ...
  • permissions — duet / stitch / share / comment / download flags
  • commerce, share, interaction, xiguaCrossPost, aiMetadata, risk
• Repeated sections: hashtags, mentions, videoTags, chapters, images, series, location

Output Example

[
  {
    "id": "7627491832754228514",
    "url": "https://www.douyin.com/jingxuan/sports?modal_id=7627491832754228514",
    "shareUrl": "https://www.iesdouyin.com/share/video/7627491832754228514/",
    "type": "video",
    "awemeType": 0,
    "awemeTypeLabel": "video",
    "horizontalType": 1,
    "horizontalTypeLabel": "horizontal",
    "language": "en",
    "transcript": "The horizontal distance between your shoulder peaks only needs to be 2 cm wider for your entire silhouette to look noticeably broader. The problem is most people train for thickness instead of width. Shoulder training really has two directions ...",
    "segments": [
      { "start": 0.0,  "end": 3.21,  "text": "The horizontal distance between your shoulder peaks only needs to be 2 cm wider," },
      { "start": 3.21, "end": 6.07,  "text": "for your entire silhouette to look noticeably broader." },
      { "start": 6.07, "end": 9.82,  "text": "The problem is most people train for thickness instead of width." }
      // ... 73 more cues
    ],
    "srtUrl": "https://api.apify.com/v2/key-value-stores/<store-id>/records/7627491832754228514.srt",
    "text": "决定肩宽的5个肌群，5招解锁3D大宽肩，视觉宽度直接+2cm 为什么你每天练推举，肩膀还是不见宽？...",
    "caption": "为什么你每天练推举，肩膀还是不见宽？...",
    "itemTitle": "决定肩宽的5个肌群，5招解锁3D大宽肩",
    "descLanguage": "zh",
    "duration": 134.305,
    "createTime": 1775913854,
    "createDate": "2026-04-11",
    "region": "",
    "city": "Henan",
    "cityCode": "411300",
    "authorMeta": {
      "id": "4297362470867675",
      "secUid": "MS4wLjABAAAAY0GGwerhkqmPWw_1UUtglG3P8btfpFqRQnWOt1_fEVsWyeuktMR3id1pChWn7XQ4",
      "name": "百里蜗牛健身",
      "username": "93576882461",
      "verified": false,
      "signature": "📖健身科普，训练不走弯路。\n🫡搞懂健身逻辑，越练越对路。",
      "avatarThumb": "https://p11.douyinpic.com/aweme/100x100/...webp",
      "totalLikesReceived": 584150,
      "country": "CN",
      "ipLocation": "河南"
    },
    "videoMeta": {
      "cover": "https://p3-sign.douyinpic.com/.../cover.jpg",
      "originCover": "https://p3-sign.douyinpic.com/.../origin.jpg",
      "height": 576,
      "width": 1024,
      "ratio": "540p",
      "duration": 134305,
      "playUrl": "https://v5-ali-northeast.douyinvod.com/.../video.mp4",
      "downloadUrl": "https://v5-ali-northeast.douyinvod.com/.../download.mp4",
      "isHdr": false,
      "hasWatermark": false
    },
    "musicMeta": {
      "id": "7627491894452472618",
      "name": "@百里蜗牛健身创作的原声",
      "author": "百里蜗牛健身",
      "isOriginal": true,
      "duration": 134
    },
    "statistics": {
      "diggCount": 180821,
      "commentCount": 1740,
      "shareCount": 21858,
      "collectCount": 116460,
      "playCount": null
    },
    "hashtags": [
      { "id": "1564018778966018", "name": "练肩" },
      { "id": "1596256922190862", "name": "肩部训练" },
      { "id": "1627603959954444", "name": "宽肩" },
      { "id": "1604211373588483", "name": "健身干货" }
    ],
    "videoTags": [
      { "name": "运动健身", "level": 1 },
      { "name": "健身", "level": 2 }
    ],
    "chapters": [
      { "startMs": 0,      "title": "Side-tilt lateral raise" },
      { "startMs": 46078,  "title": "Push-up with hip distance + lean forward" },
      { "startMs": 67973,  "title": "Hanging scapular contraction pull-ups" }
      // ... 3 more
    ],
    "permissions": {
      "canDuet": true,
      "canStitch": true,
      "canDownload": false,
      "canShare": true,
      "canComment": true
    },
    "mentions": [],
    "images": [],
    "series": null,
    "location": null,
    "errMsg": "",
    "truncatedFromSeconds": null,
    "timestamp": "2026-05-07T06:13:55+00:00"
  }
]

Advanced Usage

Custom subtitle archive across multiple runs

The default key-value store on every run gets garbage-collected after 7 days. Drop your .srt files into a named persistent store instead:

{
  "videoUrls": ["7534679152504376595", "7627491832754228514"],
  "targetLanguage": "en",
  "outputSrt": true,
  "srtKvStoreName": "douyin-subtitles-2026-q2"
}

Browse and re-download files at any time at console.apify.com/storage/key-value.

Translating one video into multiple languages

Run the actor once per target language; each run charges only for that language. The transcript stage is the expensive part — translation alone is the smaller share of the per-minute price.

{ "videoUrls": ["7534679152504376595"], "targetLanguage": "ja", "outputSrt": true }
{ "videoUrls": ["7534679152504376595"], "targetLanguage": "ko", "outputSrt": true }
{ "videoUrls": ["7534679152504376595"], "targetLanguage": "es", "outputSrt": true }

Bulk transcription from a profile or hashtag scrape

Pipe video IDs from a sibling actor straight into videoUrls. Two-step pattern:

1. Run Douyin Profile Scraper to pull a creator's recent posts.
2. Feed the id field from each post into this actor's videoUrls.

from apify_client import ApifyClient

client = ApifyClient("YOUR_APIFY_TOKEN")

# Step 1: pull every post by a creator
profile_run = client.actor("zen-studio/douyin-profile-scraper").call(
    run_input={"profileUrls": ["https://www.douyin.com/user/MS4wLj..."]}
)
posts = list(client.dataset(profile_run["defaultDatasetId"]).iterate_items())
video_ids = [p["id"] for p in posts]

# Step 2: transcribe them all in one batch
transcripts_run = client.actor("zen-studio/douyin-transcripts-scraper").call(
    run_input={
        "videoUrls": video_ids,
        "targetLanguage": "en",
        "outputSrt": True,
        "srtKvStoreName": "creator-archive",
    }
)
transcripts = list(client.dataset(transcripts_run["defaultDatasetId"]).iterate_items())

This gives you a creator's complete searchable transcript library in any language.

Want the MP4 video file too?

Just flip the toggle — this actor already downloads the video to extract audio, so saving a permanent copy is free.

{
  "videoUrls": ["https://www.douyin.com/video/7534679152504376595"],
  "targetLanguage": "en",
  "outputSrt": true,
  "shouldDownloadVideos": true,
  "shouldDownloadCovers": true,
  "videoKvStoreName": "douyin-mp4-archive"
}

The MP4 lands in your key-value store at <aweme_id>.mp4 and the cover at <aweme_id>.jpg. The dataset row carries flat top-level mp4Url and coverImageUrl strings so you can grab the file URL directly — no nested object lookup. The full videoFile / coverFile blocks are also there if you need the byte size or the original source CDN URL.

from apify_client import ApifyClient

client = ApifyClient("YOUR_APIFY_TOKEN")
run = client.actor("zen-studio/douyin-transcripts-scraper").call(
    run_input={
        "videoUrls": ["https://www.douyin.com/video/7534679152504376595"],
        "targetLanguage": "en",
        "outputSrt": True,
        "shouldDownloadVideos": True,
        "shouldDownloadCovers": True,
        "videoKvStoreName": "douyin-archive",
    }
)
for row in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(row["transcript"][:80])
    print("MP4:  ", row.get("mp4Url")        or "—")
    print("Cover:", row.get("coverImageUrl") or "—")
    print("SRT:  ", row.get("srtUrl")        or "—")

Pricing is unchanged — only the per-minute transcription rate applies. The MP4 and cover come along for free because the bandwidth is already paid for by the transcription work.

If you only want MP4s (no transcripts) the Douyin Video Scraper is the right tool: it skips the transcription step entirely and bills $4.99 / 1,000 MP4 downloads with no per-minute charge. Use it for bulk archiving; use this actor when you want transcripts and would also like a copy of the video.

Failed videos still emit a row

Every input video produces exactly one dataset row. If a video can't be loaded or transcribed (private, deleted, no audio track, etc.), the row carries a clear errMsg field:

{
  "id": "9999999999999999999",
  "url": "https://www.douyin.com/video/9999999999999999999",
  "errMsg": "Video not available",
  "transcript": "",
  "segments": [],
  "language": "zh"
}

Pricing — Pay Per Event (PPE)

Per audio-minute billing. Round-up to whole minutes — a 0:50 clip is billed as 1 minute, a 19:20 clip as 20 minutes.

Event	Per call
Original Chinese transcript per minute (zh)	$0.04
Translated transcript per minute (any non-zh language)	$0.06
.srt subtitle file export (when outputSrt: true)	$0.02

Translated rate replaces — does not stack on top of — the Chinese rate.

Cost Examples

Video	Language	SRT?	Cost
30-second clip	Chinese	No	$0.04
90-second clip	English	No	$0.12
2-minute clip	German	Yes	$0.14
10-minute video	Japanese	Yes	$0.62

Free Tier

5 audio-minutes lifetime per non-paying user, no credit card required.

For videos longer than your remaining quota, the actor processes the first N minutes that fit, ships a partial transcript, and stamps the row with an upgrade message + a truncatedFromSeconds field. Quota is only debited on successful processing — failed videos don't burn your free minutes.

Cost Optimization Tips

• Stick with targetLanguage: "zh" if your downstream pipeline already translates. The original Mandarin transcript is the cheapest and most accurate output.
• Skip outputSrt for use cases that only need the JSON segments. srtUrl is for direct video-editor / player ingestion.
• Batch 50–100 video URLs into a single run to amortize the actor-start fee.

FAQ

How accurate is the Chinese speech recognition? We use a frontier-model Mandarin ASR engine tuned for short-form social video. Quality is comparable to professional human transcripts on clean audio; degrades gracefully on heavy background music or thick regional accents.

How accurate are the translations? Two-tier LLM translation pipeline: a fast model handles the bulk, with automatic escalation to a stronger model when line-count alignment fails. Each transcript line maps 1:1 to a translated line so timestamps stay accurate.

Can I get word-level timestamps? Not currently — segments are sentence- or clause-level. Each segment carries float-second start / end precision, which is enough for subtitle display and timeline navigation.

What's the difference between this and Douyin Video Scraper? Video Scraper is for bulk MP4 / cover / slideshow archiving without transcription, billed at $4.99 / 1,000 MP4 downloads with no per-minute charge. This actor adds the spoken-content transcript, segment-level timing, optional .srt subtitle file, and translation into 50 target languages — and now also lets you save the original MP4 + cover to your key-value store as a free bonus (the bytes are already on disk for audio extraction). Pick this one when you want transcripts and would also like the video; pick Video Scraper when you only want the video.

Can I scrape transcripts at scale? Yes. Pass up to 5 URLs per run; the actor processes them concurrently. For larger batches, run multiple jobs in parallel — each is fully isolated.

Does it work with private or age-gated videos? No. The actor extracts publicly available data only. Private videos return an errMsg: "Video not available" row.

What happens with videos that have no audio? The actor detects "no audio track" cleanly and returns an empty transcript with errMsg: "Video has no audio track". You're not charged the transcription fee in this case.

Can the .srt subtitle file be uploaded to YouTube / Instagram / CapCut? Yes. The output is standard SubRip format — directly compatible with YouTube subtitle uploads, Instagram closed captions, CapCut auto-import, Premiere Pro, Final Cut Pro, DaVinci Resolve, and any other video editor that supports .srt.

How do I keep .srt files across runs? Set srtKvStoreName to any name you want (e.g. "douyin-subtitles"). Files land in a named, persistent key-value store you can browse at console.apify.com/storage/key-value. The default per-run store is wiped after 7 days; named stores live forever.

What's the free tier? 5 audio-minutes lifetime per user. For longer videos, the actor returns a partial transcript covering the first minutes that fit your quota, plus an upgrade prompt. Quota debits only on successful runs.

Does the per-minute rate include translation? Yes. The transcript_translated_minute rate ($0.06) covers both the original transcription AND the translation step — it replaces the Chinese rate, not stacks on top of it.

Support

• Bugs: Issues tab
• Features: Issues tab

Legal Compliance

Extracts publicly available data from public 抖音 video pages. Users are responsible for complying with Douyin's terms of service, copyright law, and applicable data-protection regulations (GDPR, CCPA). Downstream use of transcripts and subtitles must respect the original creator's rights — attribution is recommended when republishing.

---

Timestamped Mandarin transcripts and translated subtitles for any 抖音 (Douyin) video. 50+ languages, .srt export, full creator metadata. Frontier-model AI quality, per-minute pricing.